1 Int. J. Cancer: 115, (2005) 2005 Wiley-Liss, Inc. Predictions of mortality from pleural mesothelioma in Italy: a model based on asbestos consumption figures supports results from age-period-cohort models Alessandro Marinaccio 1 *, Fabio Montanaro 2, Marina Mastrantonio 3, Raffaella Uccelli 3, Pierluigi Altavista 3, Massimo Nesti 1, Adele Seniori Costantini 4 and Giuseppe Gorini 4 1 Epidemiology Unit, Occupational Medicine Department, National Institute for Occupational Safety and Prevention, Rome, Italy 2 Ticino Cancer Registry, Cantonal Institute of Pathology, Locarno, Switzerland 3 Unit of Toxicology and Biomedical Sciences, National Agency for New Technologies, Energy and the Environment, Rome, Italy 4 Unit of Occupational and Environmental Epidemiology, Center for Study and Prevention of Cancer, Florence, Italy Italy was the second main asbestos producer in Europe, after the Soviet Union, until the end of the 1980s, and raw asbestos was imported on a large scale until The Italian pattern of asbestos consumption lags on average about 10 years behind the United States, Australia, the United Kingdom and the Nordic countries. Measures to reduce exposure were introduced in the mid-1970s in some workplaces. In 1986, limitations were imposed on the use of crocidolite and in 1992 asbestos was definitively banned. We have used primary pleural cancer mortality figures ( ) to predict mortality from mesothelioma among Italian men in the next 30 years by age-cohort-period models and by a model based on asbestos consumption figures. The pleural cancer/mesothelioma ratio and mesothelioma misdiagnosis in the past were taken into account in the analysis. Estimated risks of birth cohorts born after 1945 decrease less quickly in Italy than in other Western countries. The findings predict a peak with about 800 mesothelioma annual deaths in the period Results estimated using ageperiod-cohort models were similar to those obtained from the asbestos consumption model Wiley-Liss, Inc. Key words: asbestos consumption; pleural cancer; mesothelioma; national data The background incidence of pleural mesothelioma (PM) (i.e., without asbestos exposure) has been estimated at about 1 2 cases per million per year, but the number of cases is much higher than expected among asbestos-exposed populations. 1,2 The disease has a very poor prognosis and survival is generally shorter than 1 year. 3 5 Mesothelioma was not well captured in mortality statistics because of the absence of a specific code in the International Classification of Disease and Related Health Problems (ICD) revisions up to ICD-9. Mortality from primary pleural cancer was therefore often used as a proxy of PM incidence despite concerns about the accuracy of death certificates. 6 8 The relation with asbestos exposure is well established and in many countries the rising pattern of PM incidence/mortality rates follows the asbestos consumption curve with a mean lag time of years. 9,10 On the basis of age-cohort models, increases in PM deaths among males are predicted in Western Europe until , 6 and recently the predictions for the Netherlands, 11 Sweden 12 and the United Kingdom 13 have been updated and corrected using new figures. Italy, Greece and former Soviet Union countries were the only asbestos producers in Europe and the other European countries almost exclusively used imported material. In the period between the end of the Second World War and the asbestos ban in 1992, Italy produced 3,748,550 tons of raw asbestos. Production rose exponentially up to the mid-1970s, reaching its peak between 1976 and 1979 at about 160,000 tons/year. Till 1987, production was more than 100,000 tons/year, then fell steeply to 0 close to 1992, the year asbestos was banned. Imports of raw asbestos peaked in , amounting to more than 77,000 tons. In the period just before the asbestos ban ( ), imports of raw asbestos (about 60,000 tons/year) exceeded domestic production. As regards the type of asbestos, Italy produced almost exclusively chrysotile; crocidolite was purchased from Australia and South Africa and amounted to 20 30% of Italian imports of raw asbestos. Reports from Denmark, Norway and Sweden show a rapid drop in asbestos imports after 1975, which were almost nil from In Finland, asbestos use reached its peak in the 1970s and then fell steeply. 17 In France, the pattern of asbestos consumption was similar to that in Italy and asbestos was banned later, in 1996, while the United Kingdom has imported virtually no asbestos since the mid-1980s. 18 In the United States, the decrease began in the mid-1970s and by 1986 the amounts were down to about 1/10 of those in the peak year. 19 Similarly, in Australia, production peaked during the decade and then dropped rapidly, stopping altogether in In Italy, measures to reduce exposure in some workplaces were introduced in the mid-1970s, limitations to the use of crocidolite were imposed in 1986, and asbestos was definitively banned in 1992, after which all further production, use and trading of any type or form of asbestos ceased. The late ban and the previous widespread use of asbestos in numerous different industrial and manufacturing sectors set Italy among the countries where the situation is particularly alarming, as shown by incidence of and mortality from mesothelioma, which in some parts of Italy are among the highest in the world. 8,21,22 The main aim of this study was to formulate projections of PM mortality in the next 30 years in Italy. Two methods were used. One was based on an age-period-cohort approach (APC): time trend in mortality rates has been analyzed and then used to predict PM deaths in the future. The second method estimated the temporal relationship between asbestos national consumption in the past and mesothelioma rates and used it for predictions. Material and methods Data concerning mortality from pleural cancer in men for the period were obtained from the Italian National Agency for New Technologies, Energy and the Environment (ENEA) s database, which collects the death certificates provided from the National Institute of Statistics (ISTAT). Pleural cancer deaths were selected according to different revisions of the ICD: the ICD-8 code (pleural cancer) was used for the period and ICD-9 code 163 (pleural cancer) for Standardized rates were calculated by the direct method using the European standard population. The distribution by age and sex of the Italian resident population from and estimates for were provided by the ISTAT. Among the 3 different scenarios available for the future population distribution, we chose the intermediate one. 23 *Correspondence to: Epidemiology Unit, Department of Occupational Medicine, National Institute for Occupational Safety and Prevention, Via Alessandria 220/E, Rome, Italy. Fax: Received 15 July 2004; Accepted after revision 19 October 2004 DOI /ijc Published online 11 January 2005 in Wiley InterScience (www. interscience.wiley.com). Publication of the International Union Against Cancer
2 PREDICTIONS OF MORTALITY 143 In consequence of the absence of a specific code for mesothelioma in the death certificates coding system until the most recent revision (ICD-10) and of the decrease of mesothelioma misdiagnosis likely occurred during the last 30 years, 2 corrections to the pleural cancer mortality data were needed to provide more reliable predictions of future mesothelioma deaths. One, proportion of recorded pleural cancers that are mesotheliomas. The number of deaths from pleural mesothelioma was estimated by multiplying the number of pleural cancer deaths by 0.73, a multiplier obtained from a study that estimated this proportion in the regional mesothelioma register of Tuscany, Italy (ARTMM). 7 Two, misdiagnosis. The increase in mesothelioma death rates recorded in the period might be partially due to the misdiagnosis in the past. 24 We allowed that the number of undiagnosed cases decreased by 5% per year. 13 The estimated 5% decrease in misdiagnosis implies that in 1970, the first year of our data series, diagnosis was about 80% complete. We generated 2 data sets, 1 applying the correction of the ratio only (data set 1) to the pleural cancer mortality data, the other applying both corrections (ratio misdiagnosis; data set 2). Age-period-cohort models The purpose of the APC models is to provide an analysis of the effects of age, cohort of birth and period of diagnosis on time trend data and to estimate birth cohorts (age and period) relative risks. Predictions have been estimated applying the relative risks to the future population. The data were organized in 13 5-year age groups (25 29 to years old), 6 5-year death periods ( to ) and year overlapping birth cohorts ( ), identified by midpoint year (that is, the birth cohort 1945 comprised those born between 1940 and 1949). The diagonals of Table I, a 2-way table of age group by time period, represent these birth cohorts. The effects of age, cohort of birth and period of diagnosis were estimated by fitting a log-linear Poisson model to the age-specific death rates of data sets 1 and 2 (Table II). 25,26 A sequence of models was fitted starting with the 1-factor age model (model A), proceeding to the 2-factor age-drift (Ad), age-period (AP) and age-cohort (AC) models, and finally to the full 3-factor age-periodcohort model, testing the statistical significance of the effect of the term added at each step. The deviance/degree of freedom ratio was used to assess the goodness of fit and a resulting value of 1.50 or less was taken as an indication of satisfactory goodness of fit. 27 The 1945 birth cohort was chosen as the cohort reference category. We tried to overcome the identifiability problem typical of the full APC models by setting to 0 the effect of the 1940 cohort adjacent to the reference point. Specifying the model in this way, the number of equations is equal to the number of parameters to be estimated in the APC full models. This choice was based on the observation that in the AC model, the 1940 and 1945 cohorts showed similar effects. The AC and APC models showed the best fit in both data sets (Table II). A statistically significant nonlinear period effect was also observed. To predict the expected number of PM deaths in the period from an AC model, the age-specific rates per birth cohort were multiplied by the projected age-specific Italian population. For the APC model, the relative risk of the last period ( ) was applied to future periods. Predictions were built using different models as follows: prediction 1, AC model, data set 1; prediction 2, APC model, data set 1; prediction 3, AC model, data set 1, with relative risks for the cohorts 1945, 1950, 1955 and 1960, respectively, 1, 1, 0.50 and 0 derived from the British mesothelioma register data; 6,24 prediction 4, AC model, data set 2; prediction 5, APC model, data set 2; prediction 6, AC model, data set 2, with the cohort relative risks used in model 3. Asbestos consumption model Data about import, export and production of asbestos in Italy were obtained from the reports published by the ISTAT, 28 the Istituto Nazionale del Commercio Estero (Foreign Trade Institute) and the Direzione Generale delle Miniere del Ministero dell Industria (General Direction of Mines, Ministry of Industry). The category raw asbestos in form of powder, fibers, rock, flock or any other form (the original definition in Italian was amianto grezzo in forma di polvere, fibre, roccia, fiocchi od altra forma ) was selected. The variable consumption has been obtained by adding to the domestic production the difference between imports and exports: C t Prod t (Imp t Exp t ). To avoid biases due to changes in the resident population, we considered amounts per capita (AC pc,t ). The best statistical relationship between annual asbestos per capita consumption data and the annual number of pleural tumor deaths for men occurred 40 years after (PPT t ) was chosen according to R 2 values, the usual index of the goodness of fit of the regression models. To estimate PM deaths from pleural tumor deaths, we applied the ratio 0.73:1, mentioned above, to this model too. Among the linear model and the logarithmic, exponential and quadratic curves estimated, the best functional relationship (that is, the one with the highest value of determination coefficient R 2 ) was PPT t a AC pc,t-40 b, where a and b 0.51 (R ). This relationship, which explains almost 80% of the observed data variability, was used to estimate the future figures of pleural cancer deaths for males. The asbestos consumption curve C t has also been used to estimate cohort risks in the general population. If we denote E as the age at first exposure and L the mean duration of exposure, 29 for the generic g cohort of birth, the relative risk (RR g ) can be estimated as the ratio of the extent of asbestos consumption from g E (exposure beginning) to g E L (exposure end) to its maximum among the cohorts: Period TABLE I PLEURAL CANCER DEATHS AND DEATH RATES ( 100,000) IN MEN BY AGE AND PERIOD, ITALY, Age group , , , , , , Age-adjusted rates ( 100,000), European population. Total 1
3 144 MARINACCIO ET AL. TABLE II AGE-PERIOD-COHORT ANALYSIS OF PLEURAL CANCER DEATHS IN ITALIAN MEN AGED 25 89, Models Deviance Degrees of freedom (df) Models to compare Change in deviance (df) p-value for changing in deviance Deviance/df ratio Data set 1 Age (A) Age drift (Ad) Ad vs A (1) Age period (AP) AP vs. Ad (4) Age cohort (AC) AC vs. Ad (16) Age cohort period (APC) APC vs. AC (4) Data set 2 Age (A) Age drift (Ad) Ad vs. A (1) Age period (AP) AP vs. Ad (4) Age cohort (AC) AC vs. Ad (16) Age cohort period (APC) APC vs. AC (4) For data set 1, pleural cancer deaths corrected using the ratio of mesothelioma to pleural cancer mortality, derived for the Tuscan mesothelioma register. For data set 2, pleural cancer deaths corrected using the above ratio and the estimate of misdiagnosis. g E L g E C t dt/max g g E g E L C t dt (1) The 1945 birth cohort was chosen as the reference category. All statistical analyses were performed with Statistical Package for Social Sciences version 10.0 (SPSS, Chicago, IL) and Stata statistical software version 8.0 (Stata, College Station, TX). Results Descriptive statistics From 1970 to 1999, 13,303 pleural cancer deaths among males have been observed in Italy, corresponding to 9,711 PM deaths, if the ratio of mesothelioma to pleural cancer mortality (0.73:1) is applied. Annual pleural cancer deaths increased from 227 in 1970 to 753 in 1999; the age-standardized rates doubled over 30 years: from 1.64 in to ,000 in (Table I). Mortality increased with age, with the highest age-specific rates in the groups older than 75 years. Age-period-cohort models Table II shows the goodness of fit for the tested models. Agecohort and age-cohort-period models fit the observed data well (the value of the deviance is very close to the degrees of freedom). The introduction of each variable improves significantly the models (p-value of change in deviance ). Results of predictions of future mesothelioma deaths for males up to 2030 are shown in Figure 1. Models 1, 2 and 3 indicate peak mortality from PM among men as, respectively, about 890 deaths/ year in , about 780 deaths/year in and about 830 deaths/year in After correction for presumed misdiagnoses, models 4 and 6 provided a lower peak (about deaths/year in ), while predictions obtained with the full APC model (model 5) gave similar results to those obtained in prediction 2, with a peak in of about 780 deaths/year. Asbestos consumption model Future PM deaths among men, estimated by the model based on asbestos consumption figures, should amount to between 810 and 830/year in the period (Fig. 2). Assuming a mean age of 20 at first asbestos exposure and a mean duration of exposure of 20 years for all the cohorts being 20 years old, on the basis of the asbestos consumption curve and according to Equation 1, risks coefficients for the cohorts 1945, 1950, 1955 and 1960 were 1, 0.97, 0.77 and 0.51, respectively. Discussion Age-period-cohort models: assumptions and limitations Pleural mesothelioma deaths have been obtained by multiplying the number of pleural cancer deaths for 0.73 in all models on the basis of a study aimed to estimate the ratio of PM to pleural cancer mortality among men in the period in Tuscany. 7 The low ratio observed in that study (0.73:1) was mainly due to the poor quality of death certificates coded ICD 163. We used the ratio 0.73, assuming that this ratio remained constant over the whole period ( ) for the whole Italy, even though it was estimated only for the period in Tuscany. Although this assumption was consistent with the findings from both a survey performed by Iwatsubo et al. 30 on mesothelioma cases diagnosed in 3 French regions in the period and an analysis performed by Bruno et al. 31 on a case list of 523 cases diagnosed in in 14 Italian regions, it could represent a limitation of the study. People aged more than 75 years have been included in the analysis in order to provide a complete scenario of mesothelioma mortality in the next years, although mesothelioma diagnosis is problematic in elderly people and misclassification with other causes of death increases in this age class. A further assumption was needed as a continuing gain in completeness of diagnosis may have occurred during the period A greater awareness among clinicians on asbestos risks, the increase of cases diagnosed with histologic analyses due to a more widespread use of invasive techniques (ultrasound or CT-guided cutting needle biopsy or thoracoscopic biopsy) and the increasing use of immunohistochemical stains in pathologic examinations determined in Italy an increase in completeness of diagnosis. In an effort to provide also estimates adjusted for temporal changes in misdiagnosis, we designed models 4, 5 and 6, even though the estimated trend of misdiagnosis that we assumed of 5% per year in accordance to Health and Safety Executive hypothesis 13 should be confirmed. As stated above, 3 projections for the Italian future population (high, medium and low hypothesis) were available. We considered the intermediate scenario as the most reliable one and estimates obtained using other scenarios did not differ significantly. The 2 APC models, 2 and 5, gave similar predictions: a peak in with about 780 annual deaths. On the other hand, the AC models (predictions 1, 3, 4 and 6) differed widely; models 1 and 3 showed a predicted peak in with about annual deaths and models 4 and 6 gave a lower one (about deaths/year) in the same period. Thus, introducing an adjustment for misdiagnosis had a striking effect on AC models. The significant cohort effect in the APC models likely reflects the different pattern of asbestos exposure among birth cohorts. The increase of completeness of diagnosis, the occupational control measures set in motion during the 1970s and 1980s, the strict regulation of asbestos use and exposure in 1991 and the total ban in 1992 most likely affect all cohorts and were seen as a nonlinear period effect. It may still be too early to confirm whether the control measures for asbestos have been effective. 16 If the median latency time is years, as it is commonly described, the
4 PREDICTIONS OF MORTALITY 145 FIGURE 1 Pleural mesothelioma deaths ( ) from data set 1 (pleural cancer deaths corrected using the ratio of mesothelioma to pleural cancer mortality derived from the Tuscan mesothelioma register) and data set 2 (pleural cancer deaths corrected using the above ratio and the estimate of misdiagnosis) and predicted ( ) deaths among men aged years, Italy, according to different age-periodcohort models. delayed period effect of asbestos regulations would have its greater effects on rates around A further critical point of predictions based on pleural cancer deaths is represented by the various choices made in the statistical modeling procedures. We used an empirical procedure to handle the problem of nonidentifiability suggested from the analysis of the related 2-term models. However, this does not mean that the nonidentifiability problem has been completely solved. The effect of specific assumptions on the risks of birth cohorts of 1945 and beyond was shown in predictions 1 and 3 for men; in model 1, the estimated risks of 1945, 1950, 1955 and 1960 birth cohort were, respectively, 1, 0.80, 0.79 and 0.55; in prediction 3, based on the assumption by Peto et al., 6 they were 1, 1, 0.50 and 0. The risks by birth cohort after 1950 in Italy may have declined less quickly than in England, reflecting the different patterns of asbestos consumption. Uncertainty remains since the risks among birth cohorts from 1945 and beyond could not be estimated with precision because of the small number of observed cases. Our long-term predictions also depend on the assumption that the death rates in men born since around 1945 will continue to increase with age as sharply as in earlier generations. In reality, the rates for this and subsequent cohorts may not continue to rise as steeply as in earlier generations, since asbestos exposure fell after The findings of APC models have to be considered according to this unavoidable but important limit. Asbestos consumption model: assumptions and limitations We used the national asbestos consumption curve as an indirect measure of exposure in the general population at aggregate level. In reality, this relationship depends on the modalities of asbestos use, work conditions, the protection measures adopted, the number of exposed subjects and the levels of environmental contamination. In Italy, besides the general limitations and the official ban, measures to reduce exposure to asbestos were first adopted during the 1970s, when the use of sprayed crocidolite in railway carriage construction plants, shipyards (construction area) and the asbestoscement industry, as well as exposure due to the recycling of jute sacks, ceased. About 17% of the incident cases in the Italian mesothelioma register (ReNaM) with available occupational information had worked in one of those sectors. 32 After that, during the 1980s, asbestos use ceased in most of the major textile plants and prevention measures were introduced in shipyards (maintenance area), in the navy and in the iron and steel industry; even the contract works of insulation removal from railway carriages and maintenance of facilities for electricity production and distribution came to an end. Altogether, those sectors employed 45% of Re- NaM cases with available occupational information. Only during the 1990s were prevention measures set up in sugar refineries, in the chemical and glass industry, in petroleum extraction and refining and vehicles maintenance and repairing (totaling about 10% of ReNaM cases). Special attention was still needed in the building sector, as Italy produced a wide number of cases and exposure occurred (and can still occur) mainly during maintenance works.
5 146 MARINACCIO ET AL. FIGURE 2 Italian raw asbestos per-capita consumption (5-year moving average; tons per 1,000,000 inhabitants), observed ( ) and predicted ( ) pleural mesothelioma deaths (PM) among men aged years old in Italy. Pleural mesothelioma deaths pleural cancer deaths As the prevention measures were set in motion at different time in different sectors with potential asbestos exposure, it is difficult to quantify their effect on the number of future mesothelioma cases, but this number might well be lower and the peak will be seen earlier than would have been estimated using exclusively deaths and asbestos consumption figures. The last assumptions in the model concern the latency (in this case, the time interval elapsing between first asbestos exposure and mesothelioma diagnosis). We assumed a latency period of 40 years. The ReNaM estimated a median latency time of 44 years based on the incident cases recorded in 1997 with an occupational exposure to asbestos. The German Mesothelioma Registry estimated the mean interval between first exposure to asbestos and onset of the disease to be 38.7 years, with a wide range (11 68). 33 Mean latency times greater than 40 years among subjects occupationally exposed to asbestos were reported by the French Mesothelioma Registry 34 and the Australian Mesothelioma Registry about cases incident in the period Many studies on wide data sets and reviews report considerable variability in the mean latency reflecting the type of exposure. 36,37 Takahashi et al. 38 and Tossavainen 39 recently observed a linear relationship between the mesothelioma incidence/mortality rate and per-capita asbestos consumption in 10 Western countries (Italy included) with different lag periods (from 10 to 25 years). We found the best fit with a nonlinear curve. Asbestos consumption data have been used also to verify the distribution of the risks, following the conceptual approach adopted by Banaei et al. 29 in the analysis of French data. We took the median age of hiring in an asbestos-exposed workplace as 20 years old, constant for each cohort. This figure proved the most reliable in an analysis of data from the ReNaM and the ARTMM, though the age of first hiring varies widely in different occupational sectors, from 18 years old for some job category as sifters and masons to years old for railwaymen and shipyard workers. Even the mean duration of exposure varies by sector and, more limitedly, by birth cohort. To simplify the analysis, it was fixed and kept constant at 20 years. The concordance between cohort risks estimated on the basis of asbestos consumption figures and those obtained by the APC model was elevated, suggesting the possibility of evaluating cohort risks and the subsequent burden of deaths using asbestos consumption when no figures are available for mortality and incidence. In conclusion, the results of our study predicted that pleural mesothelioma deaths in Italian men will level off and a peak of about 800 annual deaths will likely be reached in The estimated peak of mesothelioma cases appeared to be delayed in Italy in comparison with that observed in the United States, in the United Kingdom and in the Nordic countries, likely due to the peculiar Italian asbestos consumption curve. Asbestos consumption data, the pleural cancer/mesothelioma ratio and an estimate of mesothelioma misdiagnosis trend were included in the analysis, but the influence of preventive measures that have been adopted in Italy since 1970 cannot be quantitatively included. Finally, it is noteworthy that results estimated using mortality data were similar to those obtained from the asbestos consumption model. Acknowledgements The authors thank Roberto Pasetto of the National Institute of Health (Rome, Italy) for useful comments and discussions about these themes, Alessandra Burgio of the National Institute of Statistics (Rome, Italy) for suggestions about population data and the staff of the Foreign Trade Institute (Rome, Italy) library for help in obtaining figures on raw asbestos imports and exports.
6 PREDICTIONS OF MORTALITY McDonald JC, McDonald AD. The epidemiology of mesothelioma in historical context. Eur Resp J 1996;9: Boffetta P, Burdorf A, Goldberg M, Merler E, Siemiatycki J. Towards the coordination of European research on the carcinogenic effects of asbestos. Scand J Work Environ Health 1998;24: Spirtas R, Connelly RR, Tucker MA. Survival patterns for malignant mesothelioma: the SEER experience. Int J Cancer 1988;41: Magnani C, Viscomi S, Dalmasso P, Ivaldi C, Mirabelli D, Terracini B. Survival after pleural malignant mesothelioma: a population-based study in Italy. Tumori 2002;88: Marinaccio A, Nesti M, Regional Operational Centers. Analysis of survival of mesothelioma cases in Italian register (ReNaM). Eur J Cancer 2003;39: Peto J, Decarli A, La Vecchia C, Levi F, Negri E. The European mesothelioma epidemic. Br J Cancer 1999;79: Gorini G, Merler E, Chellini E, Crocetti E, Costantini AS. Is the ratio of pleural mesothelioma mortality to pleural cancer mortality approximately unity for Italy? considerations from the oldest regional mesothelioma register in Italy. Br J Cancer 2002;86: Montanaro F, Bray F, Gennaro V, Merler E, Tyczynski JE, Parkin DM, ENCR working group. Pleural mesothelioma incidence in Europe: evidence of some deceleration in the increasing trends. Cancer Causes Control 2003;14: McDonald JC. Health implications of environmental exposure to asbestos. Environ Health Perspect 1985;62: Selikoff IJ, Hammond EC, Seidmann H. Latency of asbestos disease among insulation workers in the United States and Canada. Cancer 1980;46: Segura O, Burdorf A, Looman C. Update of predictions of mortality from pleural mesothelioma in the Netherlands. Occup Environ Med 2003;60: Hemminki K, Li X. Mesothelioma incidence seems to have leveled off in Sweden. Int J Cancer 2003;103: Health and Safety Executive. Mesothelioma mortality in Great Britain: estimating the future burden. Health and Safety Executive, Kjaergaard J, Anderson M. Incidence rates of malignant mesothelioma in Denmark and predicted future number of cases among men. Scand J Work Environ Health 2000;26: Ulvestad B, Kjaerheim K, Moller B, Andersen A. Incidence trends of mesothelioma in Norway, Int J Cancer 2003;107: Järvholm B, Englund A, Albin M. Pleural mesothelioma in Sweden: an analysis of the incidence according to the use of asbestos. Occup Environ Med 1999;56: Karjalainen A, Pukkala E, Mattson K, Tammilehto L, Vainio H. Trends in mesothelioma incidence and occupational mesotheliomas in Finland in Scand J Work Environ Health 1997;23: Gilg Soint Ilg A, Bignon J, Valleron AJ. Estimation of the past and future burden of mortality from mesothelioma in France. Occup Environ Med 1998;55: Price B, Ware A. Mesothelioma trends in the United States: an update based on surveillance, epidemiology and end results program data for 1973 through Am J Epidemiol 2004;159: Leigh J, Davidson P, Leigh H, Berry D. Malignant mesothelioma in Australia, Am J Ind Med 2002;41: References 21. Zanetti R, Crosignani P, Rosso S, eds. Il cancro in Italia: i dati di incidenza dei registri tumori. Rome: Pensiero Scientifico Editore, Parkin DM, Whelan SL, Ferlay J, Raymond L, Young J, eds. Cancer incidence in five continents. vol. 7. Lyon: IARC, Italian Institute of Statistics. Previsioni della popolazione: anni Italian Institute of Statistics. Rome, Italy; Peto J, Hodgson JT, Matthews FE, Jones JR. Continuing increase in mesothelioma mortality in Britain. Lancet 1995;345: Clayton D, Schiffler E. Models for temporal variation in cancer rates: I, age-period and age-cohort models. Stat Med 1987;6: Clayton D, Schiffler E. Models for temporal variation in cancer rates: II, age-period-cohort models. Stat Med 1987;6: McCullagh P, Nelder JA. Generalized linear models. London: Chapman and Hall, Italian Institute of Statistics. Statistiche del commercio con l estero e statistiche della produzione industriale: vari anni. Italian Institute of Statistics. Rome, Italy, Banaei A, Auvert B, Goldberg M, Gueguen A, Luce D, Goldberg S. Future trends in mortality of French men from mesothelioma. Occup Environ Med 2000;57: Iwatsubo Y, Matrat M, Michel E, Boutin C, Galateau-Salle F, Jougla E, Bignon J, Pairon JC, Brochard P. Estimation of the incidence of pleural mesothelioma according to death certificates in France. Am J Ind Med. 2002;42: Bruno C, Comba P, Maiozzi P, Vetrugno T. Accuracy of death certification of pleural mesothelioma in Italy. Eur J Epidemiol 1996; 12: Nesti M, Marinaccio A, Chellim E. Regional Operational Centers. Malignant mesothelioma in Italy, Am J Ind Med 2004;45: Neumann V, Gunther S, Muller KM, Fischer M. Malignant mesothelioma: German mesothelioma register Int Arch Occup Environ Health 2001;74: Desoubeaux N, Bouvier V, Gervais R, Galateau-Salle F, Thibon P, Leplumey T, Herbert C, Lecherbonnier Y, Daviet JP, Letourneux M. Mésothéliomes malins en Basse-Normandie: analyse descriptive, facteurs pronostiques et survie une étude de population. Rev Epidemiol Sante Publ 2001;49: Yeung P, Rogers A, Johnson A. Distribution of mesothelioma cases in different occupational groups and industries in Australia, Appl Occup Environ Hyg 1999;14: Bianchi C, Giarelli L, Grandi G, Brollo A, Ramani L, Zuch C. Latency periods in asbestos-related mesothelioma of the pleura. Eur J Cancer Prev 1997;6: Lanphear B, Buncher C. Latent period for malignant mesothelioma of occupational origin. Occup Med 1992;34: Takahashi K, Huuskonen MS, Tossavainen A, Higashi T, Okubo T, Rantanen J. Ecological relationship between mesothelioma incidence/ mortality and asbestos consumption in ten western countries and Japan. J Occup Health 1999;41: Tossavainen A. National mesothelioma incidence and past use of asbestos. Monaldi Arch Chest Dis 2003;59:146 9.